toys
toys copied to clipboard
WojciechMula
Storage for my snippets, toy programs, etc.
Just a comment that I thought might be of interest to you: swar-utf8-length can be done with one instruction less (if we don't count creation of constants). This doesn't make...
I use simd_count_bytes for avx512 and the new architecture for apple is now arm and there is no compilation possible from intel simd to arm. So do you schedule to...
If you are measuring allocator overhead for tiny allocations like 1 byte, I'd argue that you should add the cost of a pointer to the overhead. That cost is paid...
Instead of generating the addmask by repeatedly adding to the existing addmask in a [loop](https://github.com/WojciechMula/toys/blob/master/avx512-remove-spaces/avx512vbmi.cpp#L23), isn't the final mask simply the cumulative horizontal sum of the space `mask`? Looks like...
Dear @WojciechMula, In some code of our software, we use GNU popcount like this: // Start of macros and method copied from https://github.com/kimwalisch/libpopcnt #ifdef __GNUC__ #define GNUC_PREREQ(x, y) \ (__GNUC__...
Multiply-high instructions are typically more expensive both in terms of latency and throughput than multiply instrucitons, and certain ISA's/compilers (looking at you, MSVC) do not support emitting multiply-high without a...
As the title says, the 'double custom_floor(const double x)' function actually truncates negative values, not apply the mathematical floor function, which would for instance result in -15 given -14.5 as...
hi, the [_mm_cmpestri()](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_mm_cmpestri) with `_SIDD_CMP_RANGES` also can be used for solving the hex validation. would not it be faster?
